Who Are the Most Unpredictable Pitchers in Major League Baseball?

And Should We Consider Them Some of the Best?

It was never my intention to turn my reading of “Analyzing Baseball Data with R (3rd Edition)” into a Substack series, but who am I to defy the sports analytic winds breezily guiding my everyday thoughts? I’m just a caveman. This world frightens me.

But this time, these gusts were already blowing in a direction I wanted to explore. It’s the convergence of data science, game theory and pitching. Each hurler must balance throwing his best stuff with fooling the hitter in unexpected ways. For this exercise, let’s focus on the pitchers who seem to focus heavily on being unpredictable.

It stands to reason that the most unpredictable pitchers are the ones who have the largest arsenals or variety of deliveries. But, when striking that aforementioned balance, there is selection bias in that an unusual pitch will not get thrown if it isn’t at least somewhat effective. First, let’s take Statcast data from last season and use its pitch classifications. Next, let’s only count a pitch if that pitcher threw it at least ten times. Finally, let’s count up which pitchers have the largest arsenals to choose from. It turns out, only four MLB pitchers had at least eight types to choose from:

Kansas City hurler Seth Lugo is the only pitcher with nine qualifying pitch types to his name. He only throws his four-seam fastball 24% of the time, while still going to his split finger 2.6% of the time, something he introduced the prior season but showcased slightly more often in 2024.

But, unpredictability involves more than just repertoire. A notorious strikeout pitch or a four-seam opening salvo would make any pitcher predictable even if they resorted to other pitches. If there is some indicator—like the count—that largely dictates what is thrown, hitters will know what to expect. So, which pitchers can come up with unique sequences that almost no one sees coming?

Answering this question involves some math (you’ve been warned). Shannon entropy is a way to measure the uncertainty of a variable, as part of information theory. As new information about a variable is transmitted, probabilities are communicated as well based upon how surprising or rare that information is. In other words, variables with a 50-50 chance of some binary outcome occuring are considered more unpredictable, and these results have a higher entropy. However, when we can safely gauge what will happen with an outcome, we come up with a lower entropy.

PFF has performed studies involving which defensive players are the most unpredictable when it comes to their position at the snap, but other studies have looked at the unpredictability of defensive coverage classification. In baseball, there have been studies analyzing hitter volatility using similar mathematical concepts.

Here, let’s look at when pitchers utilize their arsenals by taking the rates at which each different pitch type is thrown within each count. For instance, a pitcher who starts a plate appearance with a four-seam fastball 40% of the time would have .4 for the rate in a 0-0 count. Then, using the Shannon entropy formula, we come up with entropies for each count. Finally, we add up the entropies for all possible counts, and then standardize this sum on a scale of zero to ten. A higher entropy signifies a more unpredictable pitcher. Again, all Statcast pitch classifications are in play.

As you would expect, Seth Lugo led the way because he boasts the largest arsenal. However, just because other pitchers had eight pitch types did not mean they were the most unpredictable:

Toronto Blue Jay Chris Bassitt and New York Met Tylor Megill wound up not finishing in the top ten. Bassitt’s primary pitch is his sinker while his sweeper has a putaway rate of nearly 30%, and Megill goes to his four-seam fastball close to 50% of the time. But, veteran Yu Darvish is known for his diversity and Arizona’s Merrill Kelly, who only has six qualifying pitches, does not use any one of them more than 26% of the time.

Finally, to circle back to one of the original questions: is unpredictability all that it’s cracked up to be? In general, it is not:

Using basic Earned Run Average (ERA), the most unpredictable pitchers mentioned do have stellar box score numbers. Lugo and Darvish should have phenomenal 2025 campaigns. However, as a pitcher’s entropy goes up, often so does his ERA. Perhaps actual stuff matters a little more than fooling hitters, but a combination of both could prove devastating for the opposition.

Perhaps the next step in this exercise is to break it down even further by hitter handedness, as pitchers usually adjust their tactics based upon which box they are standing in. But, to avoid making the sample size too small in the beginning, we still see some intuitive results and an interesting way to gauge who can keep hitters off balance with the mind game.